The authors' own deposited table puts menin at 0.87 of its diagnostic-sample signal over their oncoprotein-target peaks, in the sample their text describes as having menin reduced to undetectable levels - the two statements are about different denominators, and only one of them travels
The paper says menin became undetectable in the drug-treated sample. The authors' own published spreadsheet, covering the same genes in the same sample, says menin kept 87% of its signal. Both are honest: the first describes what a genome browser picture looks like, the second is a number from a within-library calculation that cannot see a whole-cell loss at all.
The same thing, for a scientist
`p148752_152985_Menin_Signal.csv` in the authors' repository carries their own quantification of menin over 1,692 unique peak intervals in both aliquots. Over the 1,595 labelled `Oncoprotein_Target`, the median AML/ALL ratio of their `sum_norm` column is **0.871**; ENL is **0.827** and DOT1L **0.900**. Their prose says *"both menin and DOT1L were reduced to undetectable levels in the menin-inhibitor-treated AML"*. The reconciliation is the normalisation: their `sum_norm` is fragment base pairs in the interval, divided by the interval length, divided by the **total base-pair coverage of the whole library**, plus a pseudocount of 1. **A within-library share cannot represent a global loss** - if every menin molecule left chromatin and the library were then made from background, the share at any interval would be whatever background gives there, not zero. Their 'undetectable' is a statement about browser tracks and peak calls; their 0.871 is a statement about a share. Neither is wrong and **the one that gets quoted is the one in the prose**.
The three numbers in one place
| instrument | menin | ENL | DOT1L | what it measures |
|---|---|---|---|---|
authors' own sum_norm table, 1,595 oncoprotein peaks | 0.871 | 0.827 | 0.900 | fragment bp per interval bp, per library total bp, +1 |
| this beat, same intervals, fragment midpoints, CPM per kb | 0.586 | 0.449 | 0.538 | midpoint density, per library total midpoints |
| this beat, top 50 promoters defined in the drug-free aliquot | 0.037 | 0.071 | 0.040 | the same, on a focal 2 kb window instead of a 13 kb domain |
Three defensible numbers for one quantity, spanning 24-fold. The spread is entirely the interval width and the pseudocount: a 13 kb domain dilutes a focal change, and adding 1 to a small density compresses ratios toward 1. The authors' pseudocount is the single largest contributor to the difference between rows 1 and 2.
Why this is a general lesson and not a complaint
The paper's conclusion is not in doubt - the browser tracks in Figure 7a are real and the peak calls did disappear. What is at stake is what a downstream reader carries away. 'Menin is undetectable under a menin inhibitor in a patient' is a sentence that will be used to argue that menin inhibition achieves complete target disengagement in vivo. The authors' own table, in the same repository, does not support that strength of claim, and no within-library measure could. When a paper's prose and its own deposited numbers disagree in strength, record both and say which one the deposited data can bear.
What would settle it
A spike-in. There is none anywhere in GSE252378 3, and its absence is why 'undetectable', 0.871 and 0.037 can all be true of the same libraries. A spike-in-normalised menin CUT&RUN in a patient on a menin inhibitor would convert every ratio in this beat from a share into an amount, and it is the single cheapest change to the protocol that would do so.
| quantity | menin retention at oncoprotein-target peaks, 152985 / 148752 |
|---|---|
| spread fold | 23.5 |
| authors prose | both menin and DOT1L were reduced to undetectable levels in the menin-inhibitor-treated AML |
| normalisation | (interval bp coverage / interval length) / total library bp coverage * 1e9, + 1 |
| spike in | none anywhere in GSE252378 |
| general lesson | when a paper's prose and its own deposited numbers disagree in strength, record both and say which the data can bear |
| status | run |
| script | analyses/menin-under-revumenib-in-an-infant-patient-pair/run.py |
three instruments — the numbers behind the chart
{
"authors_sum_norm_1595_peaks": 0.871,
"this_beat_same_intervals_midpoint_density": 0.586,
"this_beat_top50_drugfree_promoters": 0.037
}causes of spread
[ "interval width: 13 kb median domain vs 2 kb promoter window", "the authors' +1 pseudocount compresses low-signal ratios toward 1", "coverage-in-bp vs midpoint-count as the signal definition" ]
authors own table — the numbers behind the chart
{
"MENIN": 0.871,
"ENL": 0.827,
"DOT1L": 0.9
}- An exaggerated sense of how completely a menin inhibitor clears menin from chromatin in vivo supports monotherapy and discourages combination - in a patient who relapsed on that monotherapy with the fusion transcript still expressed.
- The opposite error is available too: quoting 0.871 as 'the drug barely worked' ignores that the B-ALL clone was eliminated, which is the strongest evidence in the paper that the drug engaged its target.
- Would the authors' 0.871 survive removal of the +1 pseudocount? That is one line of their code and was not re-run here.
- How many of this corpus's menin-inhibitor occupancy claims rest on prose rather than on a deposited number, and do the deposited numbers agree with the prose?
- Does any published menin-inhibitor chromatin experiment in any system carry a spike-in? If not, every menin-displacement magnitude in the literature is a share.
the same libraries read with a different denominator
the falsifier is scored on one of these three denominators
the quantity in dispute is menin's chromatin retention
the denominator problem